[llvm] [SLP][NFC]Skip large mostly-trivial trees with up to one vector compute node (PR #194703)

[Nikita Popov via llvm-commits]

nikic wrote:

Looks like this causes a compile-time regression: https://llvm-compile-time-tracker.com/compare.php?from=81963b129e9008b0fdbaf56a423a6514ee9a3f26&to=5249e5527f48e1dd6b13c3dabed4992b299bfcb9&stat=instructions:u

libclamav_nsis_LZMADecode.c is up 14% without code size change.

Is this expected? Skipping trees doesn't sound like it should make thing more expensive...

https://github.com/llvm/llvm-project/pull/194703